WorksPhysics and Astronomy
Nuclear Mott transition with neural quantum states
Abstract
Bound clusters can dissolve in nuclear matter due to Pauli blocking, a phenomenon known as the Mott transition that is a key determinant of the nuclear abundances in heavy-ion collisions. Microscopic descriptions of this behavior require solving the in-medium Schrödinger equation. However, this remains a major computational challenge because the nuclear medium breaks Galilean invariance and couples angular-momentum channels at finite center-of-mass momentum. The emergence of weakly bound states near dissociation further places stringent demands on conventional methods. Here, we address these challenges by introducing MottNet, a momentum-space neural quantum state framework based on self-attention that directly solves the -body in-medium Schrödinger equation. Benchmarked against exact diagonalization of the in-medium deuteron, MottNet reveals substantially weaker Mott suppression for particles in hot and dense nuclear matter than do conventional calculations based on combined approximations. At high baryon densities, the predicted -particle abundance is enhanced by orders of magnitude, even after continuum-edge subtraction. This finding offers a microscopic explanation for the unexpectedly large -particle fraction inferred from a combined analysis of FOPI heavy-ion collision data and transport model calculations.